Executive Summary
A top priority of the Business Health Care Group of Wisconsin (BHCG) and its member employers is to purchase high-value health care. To that end, BHCG, with generous support from the Greater Milwaukee Business Foundation on Health (GMBFH), commissioned a study from GNS Healthcare – a leading provider of artificial intelligence (AI) applications in health care – to evaluate the value of care provided by physicians throughout Wisconsin. GNS Healthcare performed this groundbreaking study using the claims data available through the WHIO.
Wisconsin executive health care leaders gathered to attend three separate presentations in the Milwaukee, Madison, and Wausau areas on December 4 (Milwaukee & Madison) and December 5 (Wausau) to learn more about the BHCG/WHIO/GNS Healthcare Wisconsin Physician Value Study.
The primary goals of the Physician Value Study are to drive performance improvements within health systems and inform employer benefit programs design to support better health care decision-making by encouraging patient’s use of higher-value providers. As such, the results of this study will not be publicly reported.
Wisconsin’s Value Today
The Era of Cost Containment began in the 1980s with the advent of payment structure changes such as the Center for Medicare and Medicaid Services (CMS) Diagnostic Related Groups for inpatient care. Since this time, Wisconsin’s health care providers have worked diligently to improve care outcomes anticipating that higher quality would lead to lower costs. In the early 2000s, the CMS and other payers devised payment incentives based on the quality, safety, and patient experience of care across multiple care settings, continuing to expand these programs over time. In response to these and other activities, Wisconsin has experienced significant improvement in quality outcomes.
The 2018 National Health Quality and Disparities Report (NHQDR), produced annually by the Agency for Healthcare Research and Quality (AHRQ), indicates that Wisconsin has a “strong” rating based on multiple measures including outcomes in priority areas and treatment of most diseases and conditions. Additionally, care for priority populations (disparities) is similar to national averages. However, Wisconsin does not have enough data to be rated in the NHQDR on its affordability of care. Other studies have demonstrated that Wisconsin’s quality of care comes at a high price. For example, a 2018 analysis by WalletHub ranked Wisconsin ninth in health care outcomes and 47th in cost.
Purpose of the Study
A top priority of the BHCG and its member employers is to purchase high-value health care. The BHCG employer members have tested multiple strategies to reduce the growth of health care spending in Eastern Wisconsin. Examples include care management support, onsite wellness services, and benefit plan redesign, providing financial incentives for employees to choose higher-quality, lower-cost physicians. While these strategies have produced cost savings, it was unclear what the total savings opportunity might be if a benefit plan redesign was applied throughout WI or if other types of service models were implemented.
With funding from the GMBFH, the BHCG commissioned GNS Healthcare, a leading provider of artificial intelligence applications, to conduct a study using data curated from the WHIO. The WHIO, Wisconsin’s all-payer claims database (APCD), contains data on over four million insured Wisconsin lives in the 2017 data mart. Chosen for the study for its breadth of data, ability to evaluate both primary and specialty care physicians, capacity to appraise quality and cost of care simultaneously, and proven data enhancements including normalized pricing and risk-adjusted episodes of care, the WHIO data facilitated a like for like comparison of a wide range of physicians.
The study sought to answer specific questions.
- What is the quality and efficiency of each Primary Care Physician (PCP)?
- How do quality and efficiency vary across practice groups?
- What is the savings potential of moving patients to higher efficiency PCPs, improving the performance of lower-performing PCPs, or both?
- What clinical care patterns differentiate higher and lower performing PCPs?
- What is the savings potential in select specialty care areas?
The results of this study will help support health system performance improvement and inform employer benefits design for better health care decision-making.
PCP Study Design
The study used the WHIO 2017 data mart. In total, the study included 456,753 patients with a mean age of 47 years and a nearly even split of males (48%) and females (52%). The study included patients covered by commercial (41%), Medicaid (41%) and Medicare (18%) insurance who met the criteria of having both medical and pharmacy benefits for the entire study year.
Primary care providers (PCPs) with an assigned specialty of family medicine (60%), internal medicine (26%), or pediatrics (14%) were included in the study. Patients were attributed to a PCP using one of two methods. First, if a patient was assigned to a PCP in the WHIO data, that assignment was used. Second, if no assignment was present in the WHIO data, a patient was attributed to the physician that delivered the most services to the patient based on cost. After attribution, less than one percent of patients had more than one PCP in the data set.
A total of 26 common conditions (see Appendix A for conditions) were selected for performance evaluation using risk-adjusted Episode Treatment Groups (ETGs). Outlier cases were removed from the data set based on the outlier flag in the data mart. For each episode, the results of multiple common evidence-based measures (EBMs) were determined. Average normalized price was used to measure the “cost of care.”
A total of 3,760 PCPs were included in the study after requiring that an individual PCP have at least 100 total observations across the EBM measures. For each PCP, a quality of care score was calculated for each EBM. EBM-specific scores were then aggregated using a weighting methodology that took into account the number of patients the physician had treated for each condition. Finally, a ranking of 1-4 was assigned (1-outstanding performer, 4-below average performer) to each PCP based on a requirement that there be 80% confidence that the PCP was categorized correctly. This high confidence interval was selected to improve the accuracy and stability of the ratings over time.
The cost ranking was modeled using the GNS Healthcare machine learning platform REFS (Reverse Engineering Forward Simulation) to predict the cost of each patient for each disease episode. The model adjusted for potential confounders and risk factors outside of the control of the physician, such as patient age, gender, complications, number of and specific comorbidities, severity, and insurance line of business. A cost-efficiency score for each patient episode was calculated as the model’s predicted cost divided by the actual episode cost. An overall physician cost-efficiency score was calculated based on a weighted average of episode-specific cost-efficiency scores, using the number of disease-specific episodes attributed to the physician. As shown in Figure 1 below, if a PCP’s actual episode cost was equal to the predicted cost, the efficiency score was zero. When actual cost was greater than predicted cost, the cost-efficiency score was less than 0. When actual cost was less than predicted cost, the cost-efficiency score was greater than 0.
As with the quality ranking, a cost-efficiency rank was assigned to each physician (1-outstanding performer, 4-below average performer) based on a requirement that there be 80% confidence that the physician was categorized correctly. This high confidence interval was selected to improve the accuracy and stability of the ratings over time.
The results of the quality and cost ranking of PCPs included in the study are listed in Table A.
| Rank | Quality PCP Count | Cost PCP Count | Rank Name | Rank Description |
| 1 | 502 (13%) | 260 (7%) | Outstanding Performers | We are 80% confident these providers perform better than the 75th percentile |
| 2 | 133 (4%) | 1 (0%) | Good Performers | We are 80% confident these providers perform better than the 50th percentile, but not better than the 75th percentile |
| 3 | 1,806 (48%) | 2,715 (72%) | Typical Performers | We are neither 80% confident performance is better than the 75th percentile nor 80% confident performance is worse than the 50th percentile |
| 4 | 1,319 (35%) | 784 (21%) | Below Average Performers | We are 80% confident performance is worse than the 50th percentile |
Finally, the quality and cost ranking for each PCP were compared on a grid. (See Figure 2.) These results indicate that there is almost no correlation between the quality and cost of care provided by the PCPs in this study. There were only 141 PCPs that ranked better than the 80th percentile for both quality and cost-efficiency.
Specialist Study Design
In addition to the PCP evaluation, GNS Healthcare performed a cost-efficiency analysis on specialists who perform the following procedures:
• Interventional cardiologists who perform PTCA;
• Orthopedic surgeons who perform Total Hip Replacement;
• Orthopedic surgeons who perform Total Knee Replacement; and
• Obstetricians who perform Deliveries.
The same methodology used to calculate cost-efficiency of PCPs was used to calculate cost-efficiency for the specialists that perform each of these procedures. Results are shown in Table B. The ranking results for each specialist procedure is included in the report.
| Episode Treatment Group | Specialist | Procedure | # of Episodes | # of Providers |
| Ischemic Heart | Interventional Cardiologist | ≥1 PTCA | 4,047 | 114 |
| Joint degeneration, localized – thigh, hip & pelvis | Orthopedic Surgeon | ≥1 hip replacement | 3,807 | 305 |
| Joint degeneration, localized – knee & lower leg | Orthopedic Surgeon | ≥1 arthroplasty knee | 5,912 | 355 |
| Pregnancy | Obstetrician | ≥1 delivery | 17,384 | 627 |
Potential Annual Cost Savings
The GNS causal learning platform was used to perform simulations to estimate the potential cost savings that would be realized if patients were steered to providers that perform above the 50th percentile for costs or if all of the specialists examined performed similarly to their peers who performed above the 50th percentile for costs. As shown in Table C, the total (normalized) cost of care delivered by the PCPs included in this study was $1,370,000,000. The estimated 2017 potential cost saving if patients were steered to PCPs that perform above the 50th percentile for costs or if all of the PCPs we examined performed similarly to their peers who performed above the 50th percentile for cost of care was $394,500,000. The aggregate (normalized) cost of care delivered by the specialists who performed the four procedures we examined was $687,000,000. The estimated 2017 potential cost saving if patients were steered to specialists that perform above the 50th percentile for costs or if all of the specialists we examined performed similarly to their peers who performed above the 50th percentile for cost of care was $100,000,000. Of the specialist procedures, the most significant cost savings would be gained by focusing on PTCA ($43 million) and knee replacement procedures ($37 million).
| PCPs | 4 Specialist Procedures | |
| Total cost in study group | $1.37B | $687M |
| Savings by Improving Performance above 50th percentile or Steering Patients to providers above 50th percentile | $394.5M | $100M |
Key Improvement Areas to Reduce Costs
GNS Healthcare utilized the WHIO cost information to classify the costs for the 26 conditions into the following utilization categories:
• Emergency room;
• Inpatient services;
• Primary care services;
• Specialty care services;
• Laboratory services;
• Pharmacy; and
• Radiology.
The average cost in each of these utilization categories, per episode, was calculated for episodes categorized in each of the four cost ranking categories (1-4) to determine which types of utilization differed the most between cost rank categories 1 and 2. This information will hopefully help physicians and health systems focus on the components of care that offer the greatest potential cost savings. This information will be included with the provider reports.
Access to the Wisconsin Physician Value Study Results
Organizations providing direct health care services to patients can contact the Wisconsin Health Information Organization (WHIO) to learn more about how to access your organization’s results from the Wisconsin Physician Value Study. Inquiries may be addressed to Info@whio.org. Members of the BHCG, please contact Jeff Kluever at jkluever@bhcgwi.org or 262-875-3312 for more information about the Physician Value Study.